A Finite Convergence Criterion for the Discounted Optimal Control of Stochastic Logical Networks

Stochastic logical networks (SLNs) are discrete-time stochastic dynamical systems with Boolean (or multivalued) logical state variables. The discounted infinite horizon optimal control problem for SLN is addressed in this paper. By resorting to the equivalent Markov decision process description, the infinite horizon optimization problem is presented in algebraic form. Then using the increasing-dimension technique, an improved finite convergence criterion, which can find the optimal stationary policy, is derived for value iteration approach. To demonstrate the theoretical value of this approach, it is applied to the optimization problems of the human-machine game and the p53-Mdm2 gene network.